An edge-computing-based precise regulation method for aluminum alloy gradient recycling of automobiles

Abstract

An edge computing-based precise regulation method for automobile aluminum alloy gradient recycling, comprising collecting automobile aluminum alloy recycling basic parameters and preliminary regulation, also comprising building edge computing node groups to collect multi-dimensional recycling parameters, constructing dynamic regulation models to learn parameter and nonlinear mapping of gradient classification threshold, deploying gradient classification units to divide recycling gradients, and configuring real-time feedback units to adjust collection strategies and operating parameters, each module forming a closed-loop regulation system; the method solves the problem of ignoring service damage and regulation lag in existing recycling, realizes precise regulation of aluminum alloy gradient recycling, and guarantees the performance stability of recycled aluminum alloy.

Description

Technical Field

[0001] This invention relates to the field of automotive aluminum alloy recycling technology, specifically a precise control method for gradient recycling of automotive aluminum alloys based on edge computing. Background Technology

[0002] Automotive aluminum alloys are widely used in engine blocks, chassis, and body components due to their lightweight and high strength properties. The recycling and regeneration of aluminum alloys after a vehicle is scrapped is a crucial link in resource recycling. Current methods for controlling automotive aluminum alloy recycling primarily rely on aluminum alloy composition parameters for classification, analyzing these parameters through a centralized data processing center to formulate smelting control strategies. However, during service, engine blocks experience cumulative fatigue damage from prolonged cyclic loads, and chassis components exposed to the external environment suffer surface corrosion damage. This service damage alters the microstructure and mechanical properties of the aluminum alloy. Existing methods completely ignore the impact of service damage parameters on recycling gradient classification, leading to aluminum alloys of the same composition being grouped into the same recycling gradient based on their degree of service damage. This results in significant differences in the performance of regenerated products, failing to meet the requirements of high-precision components. Furthermore, centralized data processing requires transmitting all collected parameters to a remote center; data transmission delays prevent timely issuance of control commands, and adjustments to parameters such as furnace temperature and holding time during smelting lag behind actual needs, further reducing recycling accuracy. To address these inherent shortcomings of ignoring service damage parameters and control lags...

[0003] Existing technologies have not yet proposed an effective integrated solution, and there is an urgent need for a gradient recycling method for automotive aluminum alloys that integrates multi-dimensional parameters and has real-time control capabilities. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and proposes a precise control method for gradient recycling of automotive aluminum alloys based on edge computing, comprising:

[0005] Collect basic parameters during the recycling process of automotive aluminum alloys and conduct preliminary control of the recycling process;

[0006] An edge computing node cluster is established, which is used for distributed collection of multi-dimensional recycling parameters of automotive aluminum alloys;

[0007] A dynamic control model is constructed, which takes the basic parameters and the multi-dimensional recycling parameters as inputs and learns the nonlinear mapping relationship between the input parameters and the aluminum alloy gradient classification threshold.

[0008] A gradient classification unit is deployed, which classifies the recycling gradient level of aluminum alloy based on the output of the dynamic control model;

[0009] Configure a real-time feedback unit, which adjusts the collection strategy and recycling process operation parameters of the edge computing node group based on the output of the gradient classification unit;

[0010] The edge computing node cluster, dynamic control model, gradient classification unit, and real-time feedback unit form a closed-loop control system to achieve precise control of aluminum alloy gradient recycling.

[0011] Preferably, the edge computing node group includes parameter acquisition sub-nodes, data preprocessing sub-nodes, and communication sub-nodes. The parameter acquisition sub-nodes are fixedly connected to the part to be detected in the aluminum alloy recycling to acquire raw parameters. The data preprocessing sub-nodes perform noise reduction and normalization processing on the raw parameters to generate the multi-dimensional recycling parameters. The communication sub-nodes use the 5G edge communication protocol to transmit the multi-dimensional recycling parameters to the dynamic control model.

[0012] Preferably, the multi-dimensional recycling parameters include aluminum alloy composition parameters, service damage parameters, and smelting process parameters. The aluminum alloy composition parameters are the mass fractions of Si, Mg, and Cu elements in the aluminum alloy. The service damage parameters are the cumulative fatigue counts and surface corrosion depth of the aluminum alloy during use. The smelting process parameters are the furnace temperature and holding time during aluminum alloy smelting.

[0013] Preferably, the dynamic control model includes an improved deep neural network, which includes an input layer, a hidden layer, an attention mechanism layer, and an output layer. The number of neurons in the input layer is equal to the total dimension of the basic parameters, aluminum alloy composition parameters, service damage parameters, and smelting process parameters. The hidden layer includes five fully connected layers, and the neurons in the hidden layer use the ELU activation function. The attention mechanism layer assigns weights to the feature vectors output by the hidden layer. The output layer outputs the probability value of the aluminum alloy gradient classification to support the level division of the gradient classification unit.

[0014] Preferably, the dynamic control model calculates the fit between the service damage parameters and the aluminum alloy gradient classification threshold using a service damage correction coefficient. The calculation formula is: ;

[0015] In the formula: This is a service damage correction factor, dimensionless, with a value range of 0.5-1.0. The closer the value is to 1.0, the smaller the impact of service damage on gradient classification; The fatigue damage influence coefficient is determined by the fatigue strength characteristics of aluminum alloys, calibrated through fatigue life tests, and is dimensionless. The cumulative number of fatigue events is expressed in times. The fatigue accumulation index is determined by the microscopic grain boundary structure of the aluminum alloy and is dimensionless, determined by scanning electron microscopy observation and testing. The corrosion damage coefficient is determined by the corrosion resistance of the aluminum alloy and is calibrated through salt spray corrosion testing. The unit is the reciprocal of the millimeter (mm). -1 ); The surface corrosion depth is expressed in millimeters (mm). The temperature-sensitive correction factor is determined by the thermal stability of the aluminum alloy, calibrated through high and low temperature cycling tests, and is dimensionless. The furnace temperature is expressed in Kelvin (K). This is the reference temperature for aluminum alloy melting, measured in Kelvin (K).

[0016] Preferably, the dynamic control model calculates the correlation between the smelting process parameters and the aluminum alloy recycling performance through a smelting efficiency control coefficient. The calculation formula is:

[0017] ;

[0018] In the formula: This is a dimensionless coefficient for controlling smelting efficiency, ranging from 0.4 to 1.0. The closer the value is to 1.0, the better the effect of the smelting process on improving the recyclability of aluminum alloys; The heat preservation time influence coefficient is determined by the melting and diffusion characteristics of aluminum alloys, calibrated through high-temperature diffusion tests, and is dimensionless. The heat preservation time is measured in minutes (min). The time decay index is determined by the volatilization rate of elements in the aluminum alloy and is calibrated through thermogravimetric analysis. Its unit is the reciprocal of minutes (min). -1 ); The Si element proportion coefficient is determined by the strengthening effect of Si in aluminum alloys, is calibrated through mechanical property tests, and is dimensionless. The mass fraction of the Si element is dimensionless. The reference mass fraction of Si is dimensionless. The Mg element ratio coefficient is determined by the strengthening effect of Mg in aluminum alloys, calibrated through mechanical property tests, and is dimensionless. The mass fraction of the Mg element is dimensionless. The reference mass fraction of Mg is dimensionless. This is the service damage correction factor.

[0019] Preferably, the gradient classification unit determines the degree of fit between the actual state of the aluminum alloy and the target recycling gradient through a gradient matching coefficient. The calculation formula is:

[0020] ;

[0021] In the formula: This is the gradient matching degree coefficient, dimensionless, with a value range of 0.3-1.0. The closer the value is to 1.0, the better the actual state of the aluminum alloy matches the target recycling gradient; The deviation coefficient for Cu element is determined by the age hardening characteristics of Cu element in aluminum alloys, and is calibrated through aging treatment tests. It is dimensionless. The mass fraction of Cu element is dimensionless. The reference mass fraction of Cu is dimensionless. The fatigue deviation correction factor is determined by the reversibility of fatigue damage in aluminum alloys, calibrated through fatigue repair tests, and is dimensionless. This is the baseline value for the cumulative number of fatigue events, expressed in times. The cumulative number of fatigue events is expressed in times. This is the service damage correction factor; This refers to the smelting efficiency control coefficient. Preferably, the adjustment strategy of the real-time feedback unit includes parameter acquisition frequency adjustment and smelting parameter adjustment, wherein the parameter acquisition frequency is adjusted such that when the gradient matching coefficient... When the value is less than 0.6, the real-time feedback unit increases the sampling frequency of the parameter acquisition sub-node. When the gradient matching coefficient is less than 0.6, the sampling frequency of the parameter acquisition sub-node is increased. When the value is greater than 0.8, the real-time feedback unit reduces the sampling frequency of the parameter acquisition sub-node; the smelting parameters are adjusted to match the smelting efficiency control coefficient. When the value is less than 0.5, the real-time feedback unit increases the furnace temperature or extends the holding time.

[0022] Preferably, the edge computing node cluster further includes data storage sub-nodes, which are used to store the basic parameters, multi-dimensional recycling parameters, and corresponding data storage sub-nodes. , , The data storage sub-nodes adopt a distributed storage architecture, and each parameter acquisition sub-node corresponds to a local storage module. The storage period of the local storage module is consistent with the sampling period of the parameter acquisition sub-node.

[0023] Preferably, it also includes a warning unit, which monitors in real time. , and The numerical change, when Less than 0.6 Less than 0.5 or When the value is less than 0.4, the early warning unit issues a corresponding service damage warning, smelting efficiency warning, or gradient matching warning. The early warning unit transmits the warning information synchronously with the remote monitoring platform through the local audio-visual module of the edge computing node group.

[0024] Technical effects:

[0025] This invention achieves distributed acquisition of multi-dimensional parameters through edge computing node clusters, solving the problem of lagging centralized data processing; the dynamic control model introduces a service damage correction coefficient to compensate for the shortcomings of existing methods that ignore service damage; the three coefficient formulas work together to realize the progressive influence between parameters, improving the accuracy of gradient classification; the closed-loop control system and early warning unit ensure the stability of the recycling process, ultimately achieving a consistent improvement in the performance of aluminum alloy recycling. Attached Figure Description

[0026] Figure 1 This is a flowchart of the precise control method for gradient recycling of automotive aluminum alloys based on edge computing, as described in this application. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0028] Traditional technical solutions have the following technical problems: existing automotive aluminum alloy recycling control relies solely on composition parameters, completely ignoring the impact of service damage on aluminum alloy performance, and the centralized data processing results in delayed control commands and a lack of closed-loop control mechanism, making it impossible to achieve the accuracy of gradient recycling.

[0029] Based on this, this embodiment provides a method for precise control of gradient recycling of automotive aluminum alloys based on edge computing, including:

[0030] S1: Collect basic parameters during the recycling process of automotive aluminum alloys and conduct preliminary control of the recycling process;

[0031] S2: Establish an edge computing node cluster, which is used for distributed collection of multi-dimensional recycling parameters of automotive aluminum alloys;

[0032] S3: Construct a dynamic control model, which takes the basic parameters and the multi-dimensional recycling parameters as inputs and learns the nonlinear mapping relationship between the input parameters and the aluminum alloy gradient classification threshold.

[0033] S4: Deploy a gradient classification unit, which classifies the recycling gradient level of aluminum alloy based on the output of the dynamic control model;

[0034] S5: Configure a real-time feedback unit, which adjusts the collection strategy and recycling process operation parameters of the edge computing node group based on the output of the gradient classification unit;

[0035] S6: The edge computing node group, dynamic control model, gradient classification unit and real-time feedback unit form a closed-loop control system to achieve precise control of aluminum alloy gradient recycling.

[0036] The core of this technical solution lies in constructing a closed-loop system for data acquisition, modeling, classification, and feedback. The edge computing node cluster adopts a distributed architecture, unlike traditional centralized data acquisition. Multiple sub-nodes can be deployed at the recycling site, such as in dismantling workshops or near smelting furnaces, to directly collect parameters closely reflecting the actual state of the aluminum alloy, reducing data transmission distance and thus lowering latency. The dynamic control model does not simply fit the relationship between composition parameters and classification thresholds; instead, it integrates basic parameters with multi-dimensional recycling parameters, focusing on learning the nonlinear correlation between easily overlooked parameters such as service damage and the smelting process and gradient classification thresholds. For example, even if the composition of a severely damaged aluminum alloy meets the standards, it still needs to be classified into a lower recycling gradient. The gradient classification unit determines the level based on the probability value output by the model, rather than a single threshold, improving classification flexibility.

[0037] The real-time feedback unit breaks the traditional one-way process of data acquisition and processing. It adjusts the acquisition frequency and smelting parameters in reverse according to the classification results, so that the modules form a closed loop of coordinated linkage, ensuring timely and accurate control.

[0038] The technical effects achieved by the above embodiments include: solving the problems of ignoring service damage and control lag, realizing multi-module collaboration through a closed-loop system, ensuring the accuracy of aluminum alloy gradient recycling, and providing support for stable regeneration performance.

[0039] Traditional technical solutions have the following technical problems: existing edge computing node groups do not have a clear division of sub-node functions, parameter acquisition and data processing are mixed, and there is a lack of dedicated communication protocols, resulting in insufficient data noise reduction and high transmission latency, which affects the efficiency of subsequent regulation.

[0040] Based on this, the edge computing node group includes parameter acquisition sub-nodes, data preprocessing sub-nodes, and communication sub-nodes. The parameter acquisition sub-nodes are fixedly connected to the part to be detected in the aluminum alloy recycling to collect raw parameters. The data preprocessing sub-nodes perform noise reduction and normalization processing on the raw parameters to generate the multi-dimensional recycling parameters. The communication sub-nodes use the 5G edge communication protocol to transmit the multi-dimensional recycling parameters to the dynamic control model.

[0041] This technical solution clarifies the function of each sub-node through a layered design, ensuring a clear and implementable data processing flow. Parameter acquisition sub-nodes need to be specifically fixed to the parts to be tested. For example, when acquiring service damage parameters, the sensor needs to be fixed at stress concentration points in the engine block or in areas prone to chassis corrosion to ensure that the raw parameters accurately reflect the condition of the aluminum alloy and avoid data distortion due to improper installation.

[0042] The data preprocessing sub-node has a clearly defined processing flow. Noise reduction uses mature algorithms such as wavelet thresholding or mean filtering to remove noise generated during the acquisition process due to equipment vibration and electromagnetic interference. Normalization uses Min-Max or Z-Score methods to transform parameters of different dimensions to a unified range, avoiding the impact of numerical range differences on model training accuracy. Only the processed data is used as multi-dimensional recovery parameters. The communication sub-node uses the 5G edge communication protocol. Unlike traditional Wi-Fi or 4G, its edge computing characteristics allow some data processing to be completed at the edge node close to the acquisition end, transmitting only valid parameters. At the same time, the high bandwidth and low latency of 5G ensure that multi-dimensional recovery parameters can be quickly transmitted to the dynamic control model, avoiding control lag caused by data congestion.

[0043] The technical effects achieved by the above embodiments include: clarifying the functional division of sub-nodes, improving the quality of raw data, reducing parameter transmission delay, and providing timely and accurate input data for dynamic control models.

[0044] Traditional technical solutions have the following technical problems: the existing multi-dimensional recycling parameters are vaguely defined, and the specific parameter categories are not clearly defined. They are only mentioned in general terms, which leads to confusion in the collection direction and fails to cover the key influencing factors of aluminum alloy recycling, thus affecting the targeted control.

[0045] Based on this, the multi-dimensional recycling parameters include aluminum alloy composition parameters, service damage parameters, and smelting process parameters. The aluminum alloy composition parameters are the mass fractions of Si, Mg, and Cu elements in the aluminum alloy. The service damage parameters are the cumulative fatigue counts and surface corrosion depth of the aluminum alloy during use. The smelting process parameters are the furnace temperature and holding time during aluminum alloy smelting.

[0046] This technical solution provides a clear basis for data collection by defining parameter categories and specific indicators, ensuring that multi-dimensional parameters comprehensively cover the entire recycling process. The aluminum alloy composition parameters focus on Si, Mg, and Cu elements, as these three elements are core alloying elements in commonly used 6-series aluminum alloys for automobiles, directly affecting the mechanical properties of aluminum alloys such as strength and hardness. For example, Si can improve casting performance. and form To achieve age-based strengthening, Cu can further improve strength, and collecting its mass fraction can provide a basis for judging the basic properties of aluminum alloys. Service damage parameters address the actual use scenarios of automotive aluminum alloys. Components such as engine blocks and chassis will experience cumulative fatigue damage due to cyclical loads during service. Exposure to coastal or mountainous environments will lead to surface corrosion. The cumulative fatigue count reflects the degree of fatigue damage, and the surface corrosion depth reflects the degree of corrosion damage. These two parameters can supplement the impact of service wear that cannot be reflected by composition parameters. Melting process parameters cover key regeneration stages. Furnace temperature directly affects the melting quality and element volatilization of aluminum alloys, while holding time affects compositional uniformity. Collecting these two parameters allows for real-time monitoring of the melting state, providing data support for subsequent adjustments to melting strategies. These three types of parameters correspond to original performance, service wear, and the regeneration process, forming a complete parameter collection system.

[0047] The technical effects achieved by the above embodiments include: clarifying the specific range and indicators of multi-dimensional recovery parameters, avoiding confusion in the collection direction, and providing comprehensive and accurate basic data for dynamic control models and subsequent gradient classification.

[0048] Traditional technical solutions have the following technical problems: existing dynamic control models mostly use simple neural network structures without attention mechanisms, which cannot highlight the impact of core parameters such as service damage and key components. In addition, the hidden layer activation function uses the traditional ReLU, which is prone to neuron death. As a result, the model cannot effectively learn the nonlinear mapping relationship between multiple parameters, and the gradient classification prediction accuracy is low.

[0049] Based on this, the dynamic control model includes an improved deep neural network, which includes an input layer, a hidden layer, an attention mechanism layer, and an output layer. The number of neurons in the input layer is equal to the total dimension of the basic parameters, aluminum alloy composition parameters, service damage parameters, and smelting process parameters. The hidden layer includes five fully connected layers, and the neurons in the hidden layer use the ELU activation function. The attention mechanism layer assigns weights to the feature vectors output by the hidden layer. The output layer outputs the probability value of the aluminum alloy gradient classification to support the level division of the gradient classification unit.

[0050] This technical solution enhances the model's attention to key parameters and its learning ability by optimizing the neural network structure and introducing an attention mechanism. The number of neurons in the input layer must strictly match the total dimension of the parameters. For example, if there are 2 basic parameters, 3 component parameters, 2 service damage parameters, and 2 smelting process parameters, with a total dimension of 9, then the number of neurons in the input layer should be 9, ensuring that each key parameter can be independently input into the model. Five fully connected layers are used in the hidden layers. The number of layers is designed based on parameter complexity, decreasing from 128, 64, 32, 16 to 8 neurons, which conforms to the logic of feature extraction from coarse to fine, and can gradually uncover deep correlations between parameters. The ELU activation function is selected, which retains a smaller gradient when the input is negative. To avoid the neuron death problem caused by ReLU's gradient reaching zero on negative inputs, this approach is particularly suitable for scenarios where service damage parameters may exhibit negative fluctuations. The attention mechanism layer is a core improvement. By dividing the feature vector output from the hidden layer into a query vector Q, a key vector K, and a value vector V, attention scores are calculated and normalized to obtain weights. Higher weights are assigned to feature vectors corresponding to core parameters such as service damage and key components. For example, the weights of features corresponding to cumulative fatigue counts and Si element quality scores are increased, ensuring that the model focuses on learning these parameters that significantly affect gradient classification. The output layer uses the Softmax activation function, outputting three probability values ​​for recovering gradient levels. The level with the highest probability is the final output of the gradient classification unit, providing a quantitative basis for level division.

[0051] The technical effects achieved by the above embodiments include: increasing the model's focus on core parameters, avoiding neuron death problems, enhancing the model's ability to learn multi-parameter nonlinear mapping relationships, and improving the prediction accuracy of aluminum alloy gradient classification.

[0052] Traditional technical solutions have the following technical problems: the impact of service damage on gradient classification is not quantified, and the existence of damage is determined only by qualitative description. It is impossible to accurately calculate the compatibility between the degree of damage and the gradient classification threshold of aluminum alloy, resulting in a lack of scientific coefficient support for gradient classification and strong subjectivity in the classification results.

[0053] Based on this, the dynamic control model calculates the fit between the service damage parameters and the aluminum alloy gradient classification threshold using a service damage correction coefficient. The calculation formula is: ;

[0054] In the formula: This is a service damage correction factor, dimensionless, with a value range of 0.5-1.0. The closer the value is to 1.0, the smaller the impact of service damage on gradient classification; The fatigue damage influence coefficient is determined by the fatigue strength characteristics of aluminum alloys, calibrated through fatigue life tests, and is dimensionless. The cumulative number of fatigue events is expressed in times. The fatigue accumulation index is determined by the microscopic grain boundary structure of the aluminum alloy and is dimensionless, determined by scanning electron microscopy observation and testing. The corrosion damage coefficient is determined by the corrosion resistance of the aluminum alloy and is calibrated through salt spray corrosion testing. The unit is the reciprocal of the millimeter (mm). -1 ); The surface corrosion depth is expressed in millimeters (mm). The temperature-sensitive correction factor is determined by the thermal stability of the aluminum alloy, calibrated through high and low temperature cycling tests, and is dimensionless. The furnace temperature is expressed in Kelvin (K). This is the reference temperature for aluminum alloy melting, measured in Kelvin (K).

[0055] This technical solution quantifies the impact of service damage through formulas. Each parameter and calculation item has a clear physical meaning and calibration method, ensuring that the coefficients are verifiable and implementable. As a dimensionless coefficient, its value range directly reflects the degree of damage. A value close to 1.0 indicates small damage, and the gradient classification can more closely match the level corresponding to the original component; a value close to 0.5 indicates large damage, and the gradient level needs to be lowered. (Formula Part 1) Corresponding to the effects of fatigue damage, Calibration is achieved through fatigue testing, such as applying alternating loads to Q6061 aluminum alloy samples and fitting the relationship between fatigue cycles and strength decay. , This refers to the actual cumulative number of fatigue cycles collected, such as after 5 years of service for the engine block. Second-rate, Grain boundary cracks were observed using an electron microscope, such as... This reflects the rate of cumulative fatigue damage with each repetition; this value increases with... An increase followed by a decrease indicates more severe fatigue damage. The smaller. Part Two Corrosion damage impact, Calibration is achieved through salt spray testing, such as immersion in a 5% NaCl solution. mm -1 , The actual corrosion depth, such as in chassis components. mm, in exponential form, ensures that this value decreases smoothly as the corrosion depth increases, consistent with the progressive characteristics of corrosion damage. Part Three Correction to the corresponding melting temperature, Take the melting point of the aluminum alloy, such as 6-series aluminum alloys. K, when Higher than At this point, the value is slightly greater than 1, indicating that appropriate heating can alleviate some of the damage effects, such as softening fatigue cracks. Calibration through high and low temperature cycling tests (e.g.) ).

[0056] Multiplying the three parts together yields... This comprehensively reflects the synergistic effects of fatigue, corrosion, and temperature on service damage. The technical effects achieved by the above embodiments include: quantifying the adaptation relationship between service damage and gradient classification thresholds, providing scientific coefficient support for dynamic control models, reducing the subjectivity of gradient classification, and improving classification accuracy.

[0057] Traditional technical solutions have the following technical problems: they fail to establish the correlation between smelting process parameters and aluminum alloy regeneration performance, making it impossible to determine the specific effects of adjusting parameters such as furnace temperature and holding time on regeneration performance. Furthermore, they fail to consider the early impact of service damage, leading to blind adjustments of smelting parameters and an inability to achieve coordinated control of "damage repair and performance improvement".

[0058] Based on this, the dynamic control model calculates the correlation between the smelting process parameters and the aluminum alloy recycling performance through a smelting efficiency control coefficient. The calculation formula is:

[0059] ;

[0060] In the formula: This is a dimensionless coefficient for controlling smelting efficiency, ranging from 0.4 to 1.0. The closer the value is to 1.0, the better the effect of the smelting process on improving the recyclability of aluminum alloys; The heat preservation time influence coefficient is determined by the melting and diffusion characteristics of aluminum alloys, calibrated through high-temperature diffusion tests, and is dimensionless. The heat preservation time is measured in minutes (min). The time decay index is determined by the volatilization rate of elements in the aluminum alloy and is calibrated through thermogravimetric analysis. Its unit is the reciprocal of minutes (min). -1 ); The Si element proportion coefficient is determined by the strengthening effect of Si in aluminum alloys, is calibrated through mechanical property tests, and is dimensionless. The mass fraction of the Si element is dimensionless. The reference mass fraction of Si is dimensionless. The Mg element ratio coefficient is determined by the strengthening effect of Mg in aluminum alloys, calibrated through mechanical property tests, and is dimensionless. The mass fraction of the Mg element is dimensionless. The reference mass fraction of Mg is dimensionless. This is the service damage correction coefficient. This technical solution establishes a synergistic relationship between smelting parameters, composition parameters, and service damage through a formula. Each calculation item corresponds to the specific impact of the smelting process on regeneration performance, ensuring that the coefficient can guide actual control. As a dimensionless coefficient, a value close to 1.0 indicates suitable smelting parameters and a significant improvement in regeneration performance; a value close to 0.4 indicates that the smelting parameters need adjustment. In the formula... The introduction of this concept is key, reflecting that smelting control must be based on service damage, avoiding blind adjustments that ignore early-stage damage. Part One The corresponding insulation time has an impact. Calibration is achieved through high-temperature diffusion tests, such as measuring component uniformity at 950K. , This refers to the actual heat preservation time. This value is obtained by measuring elemental volatilization through thermogravimetric analysis; it varies with... The increase followed by a decrease reflects the characteristics of uneven composition when the heat preservation time is too short and element volatilization when it is too long.

[0061] Part Two The influence of corresponding component deviations, , Take the standard value of 6-series aluminum alloy, such as , , , Calibration is achieved through tensile testing, such as , ,when , When the value approaches the baseline, this value is close to 1, indicating that smelting achieves the best performance improvement when the composition meets the standards. (The three parts are related to...) Multiply to get It comprehensively reflects the synergistic effects of service damage, holding time, and composition deviation on smelting efficiency.

[0062] The technical effects achieved by the above embodiments include: establishing the correlation between the smelting process and regeneration performance, realizing the coordinated control of service damage, smelting parameters, and composition parameters, providing a scientific basis for adjusting smelting parameters, and improving the stability of aluminum alloy regeneration performance.

[0063] Traditional technical solutions have the following technical problems: they do not comprehensively assess the degree of fit between the actual state of the aluminum alloy and the target recycling gradient, and evaluate the classification results based on a single parameter, which cannot fully measure the rationality of the gradient division and is prone to classification bias.

[0064] Based on this, the gradient classification unit determines the degree of fit between the actual state of the aluminum alloy and the target recycling gradient through a gradient matching coefficient. The calculation formula is:

[0065] ;

[0066] In the formula: This is the gradient matching degree coefficient, dimensionless, with a value range of 0.3-1.0. The closer the value is to 1.0, the better the actual state of the aluminum alloy matches the target recycling gradient; The deviation coefficient for Cu element is determined by the age hardening characteristics of Cu element in aluminum alloys, and is calibrated through aging treatment tests. It is dimensionless. The mass fraction of Cu element is dimensionless. The reference mass fraction of Cu is dimensionless. The fatigue deviation correction factor is determined by the reversibility of fatigue damage in aluminum alloys, calibrated through fatigue repair tests, and is dimensionless. This is the baseline value for the cumulative number of fatigue events, expressed in times. The cumulative number of fatigue events is expressed in times. This is the service damage correction factor; This refers to the smelting efficiency control coefficient. This technical solution comprehensively evaluates the matching between the actual state of the aluminum alloy and the target gradient by integrating multiple dimensions through a formula. Each parameter corresponds to a key influencing factor, ensuring the objectivity of the evaluation results. As a dimensionless coefficient, a value close to 1.0 indicates that the current state of the aluminum alloy perfectly matches the target gradient and no adjustment is needed; a value close to 0.3 indicates a low degree of fit and the need to re-divide the gradient or adjust the parameters.

[0067] in the formula , The introduction of this concept reflects that the initial impact of service damage and smelting efficiency has been considered, avoiding redundant calculations. Part 1 Corresponding to the influence of Cu element deviation, Cu element can improve the age hardening effect of aluminum alloys. Take the standard value of 6-series aluminum alloy. Calibration is achieved through aging tests, such as aging at 120℃ for 24 hours. The absolute value ensures that both positive and negative deviations will reduce the matching degree, while the exponential form shows that the larger the deviation, the more significant the decrease in matching degree. Part Two Corresponding fatigue deviation correction, Take the fatigue life benchmark value of aluminum alloy, such as Second-rate, Calibration is achieved through fatigue repair tests, such as low-temperature annealing. ,when When this value is greater than 1, indicating relatively low fatigue damage, the matching degree can be appropriately increased; when... When this value is less than 1, it indicates a decreased matching degree when the damage is severe. Multiplying the four parts together yields... It comprehensively reflects the synergistic effects of service damage, smelting efficiency, Cu element deviation, and fatigue deviation on gradient matching.

[0068] The technical effects achieved by the above embodiments include: comprehensively evaluating the degree of fit between the actual state of the aluminum alloy and the target gradient, avoiding the bias of single parameter evaluation, providing a quantitative basis for judging the rationality of gradient classification, and improving the accuracy of recycling gradient division.

[0069] Traditional technical solutions have the following technical problems: existing real-time feedback units lack clear adjustment strategies and have not formulated specific adjustment rules based on key coefficient values, such as the adjustment thresholds and actions for acquisition frequency and smelting parameters, resulting in blind feedback adjustments and an inability to adapt to gradient classification requirements in a timely manner.

[0070] Based on this, the adjustment strategy of the real-time feedback unit includes parameter acquisition frequency adjustment and smelting parameter adjustment, wherein the parameter acquisition frequency is adjusted when the gradient matching coefficient... When the value is less than 0.6, the real-time feedback unit increases the sampling frequency of the parameter acquisition sub-node. When the gradient matching coefficient is less than 0.6, the sampling frequency of the parameter acquisition sub-node is increased. When the value is greater than 0.8, the real-time feedback unit reduces the sampling frequency of the parameter acquisition sub-node; the smelting parameters are adjusted to match the smelting efficiency control coefficient. When the value is less than 0.5, the real-time feedback unit increases the furnace temperature or extends the holding time.

[0071] This technical solution makes real-time feedback operable by clearly defining adjustment thresholds and corresponding actions, ensuring that adjustments accurately adapt to the current state. Parameter acquisition frequency adjustment revolves around the gradient matching coefficient. Expand This indicates that the actual state of the aluminum alloy does not match the target gradient well, and there may be parameter fluctuations or key changes that have not been captured. Increasing the sampling frequency can increase the data acquisition density, detect parameter change trends in a timely manner, and provide more data support for reclassification or adjustment. This indicates high compatibility and stable status. Reducing the sampling frequency can decrease data redundancy, lower edge node power consumption, and extend equipment battery life. Adjusting smelting parameters affects the smelting efficiency control coefficient. , This indicates that the current smelting parameters are insufficient to effectively improve regeneration performance and need to be adjusted based on the actual situation. The adjustment strategies are all based on key coefficients calculated in the early stages, ensuring that each action has a clear trigger condition and avoiding blind operation.

[0072] The technical effects achieved by the above embodiments include: clarifying the adjustment rules and trigger thresholds for real-time feedback, avoiding blind adjustments, achieving precise matching between the acquisition frequency and smelting parameters, and improving the timeliness and effectiveness of closed-loop control.

[0073] Traditional technical solutions have the following technical problems: existing data storage does not adopt a distributed architecture, and the data is centrally stored in a remote center, which is prone to data loss due to network failures. In addition, the storage period is not synchronized with the sampling period, and it is impossible to correspond to the parameters and coefficient values ​​of each sampling, making it difficult to support the iterative optimization of the subsequent dynamic control model.

[0074] Based on this, the edge computing node cluster also includes data storage sub-nodes, which are used to store the basic parameters, multi-dimensional recycling parameters, and corresponding data. , , The data storage sub-nodes adopt a distributed storage architecture, and each parameter acquisition sub-node corresponds to a local storage module. The storage period of the local storage module is consistent with the sampling period of the parameter acquisition sub-node.

[0075] The storage period is strictly synchronized with the sampling period. For example, if the parameter acquisition sub-node has a sampling period of 10 seconds per sampling, the local storage module also stores one data entry every 10 seconds. Each data entry includes a timestamp to ensure that the parameters and coefficient values ​​correspond one-to-one and to avoid data corruption caused by asynchronous periods. The local storage module also has an automatic cleanup function. When storage space is insufficient, it deletes the oldest data in chronological order to ensure the continuous operation of the storage system.

[0076] The technical effects achieved by the above embodiments include: avoiding data loss, ensuring the correlation between parameters and coefficient values, and providing complete and accurate historical data support for the iterative optimization of the dynamic control model.

[0077] Traditional technical solutions have the following technical problems: they lack early warning mechanisms for service damage, smelting efficiency, and gradient matching, making it impossible to detect parameter anomalies in a timely manner. This can easily lead to substandard performance or safety hazards during the recycling process. Furthermore, the early warning information is transmitted in a single way, making it difficult for managers to obtain it in a timely manner.

[0078] Based on this, it also includes the deployment of an early warning unit, which monitors in real time. , and The numerical change, when Less than 0.6 Less than 0.5 or When the value is less than 0.4, the early warning unit issues a corresponding service damage warning, smelting efficiency warning, or gradient matching warning. The early warning unit transmits the warning information synchronously with the remote monitoring platform through the local audio-visual module of the edge computing node group.

[0079] Early warning information is transmitted synchronously in two ways: local audio-visual modules, such as LED lights and buzzers equipped on edge nodes, where service damage warnings correspond to flashing red LEDs and intermittent buzzer sounds, smelting efficiency warnings correspond to flashing yellow LEDs and continuous buzzer sounds, and gradient matching warnings correspond to alternating flashing red and yellow LEDs and a continuous buzzer sound, facilitating timely detection by on-site operators; and remote monitoring platforms, such as management platforms built on Alibaba Cloud, push warning information via SMS and an app, including the warning type, the number of the affected component, the current coefficient value, and the collection time, ensuring that management personnel can remotely monitor abnormal situations in real time. The warning unit also has a manual reset function; after the fault is handled, the operator can reset the warning via a local button, and the system records the processing time and result, forming a closed-loop warning system.

[0080] The technical effects achieved by the above embodiments include: timely detection of service damage, smelting efficiency, and gradient matching anomalies; transmission of early warning information through multiple channels; ensuring the safety and stability of the recycling process; and reducing the risk of substandard performance.

[0081] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for precise control of gradient recycling of automotive aluminum alloys based on edge computing, characterized in that, include: Collect basic parameters during the recycling process of automotive aluminum alloys and conduct preliminary control of the recycling process; An edge computing node cluster is established, which is used for distributed collection of multi-dimensional recycling parameters of automotive aluminum alloys; A dynamic control model is constructed, which takes the basic parameters and the multi-dimensional recycling parameters as inputs and learns the nonlinear mapping relationship between the input parameters and the aluminum alloy gradient classification threshold. A gradient classification unit is deployed, which classifies the recycling gradient level of aluminum alloy based on the output of the dynamic control model; Configure a real-time feedback unit, which adjusts the collection strategy and recycling process operation parameters of the edge computing node group based on the output of the gradient classification unit; The edge computing node group, dynamic control model, gradient classification unit and real-time feedback unit form a closed-loop control system to realize the control of aluminum alloy gradient recycling; The edge computing node group includes parameter acquisition sub-nodes, data preprocessing sub-nodes and communication sub-nodes. The parameter acquisition sub-nodes are fixedly connected to the detection part of the aluminum alloy recycling to collect raw parameters. The data preprocessing sub-nodes perform noise reduction and normalization on the raw parameters to generate the multi-dimensional recycling parameters. The communication sub-nodes use the 5G edge communication protocol to transmit the multi-dimensional recycling parameters to the dynamic control model. The multi-dimensional recycling parameters include aluminum alloy composition parameters, service damage parameters, and smelting process parameters. The aluminum alloy composition parameters are the aluminum alloy... The mass fraction of the elements; the service damage parameters are the cumulative fatigue number and surface corrosion depth of the aluminum alloy during use; and the smelting process parameters are the furnace temperature and holding time during aluminum alloy smelting. The dynamic control model includes an improved deep neural network, which comprises an input layer, a hidden layer, an attention mechanism layer, and an output layer. The number of neurons in the input layer is equal to the total dimension of the basic parameters, aluminum alloy composition parameters, service damage parameters, and smelting process parameters. The hidden layer comprises five fully connected layers, and the neurons in the hidden layer use the ELU activation function. The attention mechanism layer assigns weights to the feature vectors output by the hidden layer. The output layer outputs the probability value of the aluminum alloy gradient classification to support the level division of the gradient classification unit. The dynamic control model calculates the fit between the service damage parameters and the aluminum alloy gradient classification threshold using a service damage correction coefficient. The calculation formula is: ； In the formula: This is a service damage correction factor, dimensionless, with a value range of 0.5-1.

0. The closer the value is to 1.0, the smaller the impact of service damage on gradient classification; The fatigue damage influence coefficient is determined by the fatigue strength characteristics of aluminum alloys, calibrated through fatigue life tests, and is dimensionless. The cumulative number of fatigue events is expressed in times. The fatigue accumulation index is determined by the microscopic grain boundary structure of the aluminum alloy and is dimensionless, determined by scanning electron microscopy observation and testing. The corrosion damage coefficient is determined by the corrosion resistance of the aluminum alloy and is calibrated through salt spray corrosion testing. The unit is the reciprocal of the millimeter value in mm. -1 ; The surface corrosion depth is expressed in millimeters (mm). The temperature-sensitive correction factor is determined by the thermal stability of the aluminum alloy, calibrated through high and low temperature cycling tests, and is dimensionless. The furnace temperature is expressed in Kelvin (K). This is the reference temperature for aluminum alloy melting, measured in Kelvin (K).

2. The method for precise control of gradient recycling of automotive aluminum alloys based on edge computing according to claim 1, characterized in that, The dynamic control model calculates the correlation between the smelting process parameters and the aluminum alloy recycling performance through a smelting efficiency control coefficient. The calculation formula is: ; In the formula: This is a dimensionless coefficient for controlling smelting efficiency, ranging from 0.4 to 1.

0. The closer the value is to 1.0, the better the effect of the smelting process on improving the recyclability of aluminum alloys; The heat preservation time influence coefficient is determined by the melting and diffusion characteristics of aluminum alloys, calibrated through high-temperature diffusion tests, and is dimensionless. The heat preservation time is measured in minutes (min). The time decay index is determined by the volatilization rate of elements in the aluminum alloy and is calibrated through thermogravimetric analysis. Its unit is the reciprocal of minutes (min). -1 ); The Si element proportion coefficient is determined by the strengthening effect of Si in aluminum alloys, is calibrated through mechanical property tests, and is dimensionless. The mass fraction of the Si element is dimensionless. The reference mass fraction of Si is dimensionless. The Mg element ratio coefficient is determined by the strengthening effect of Mg in aluminum alloys, calibrated through mechanical property tests, and is dimensionless. The mass fraction of the Mg element is dimensionless. The reference mass fraction of Mg is dimensionless. This is the service damage correction factor.

3. The method for precise control of automotive aluminum alloy gradient recycling based on edge computing according to claim 2, characterized in that, The gradient classification unit determines the degree of fit between the actual state of the aluminum alloy and the target recycling gradient by using a gradient matching coefficient. The calculation formula is: ； In the formula: This is the gradient matching degree coefficient, dimensionless, with a value range of 0.3-1.

0. The closer the value is to 1.0, the better the actual state of the aluminum alloy matches the target recycling gradient; The deviation coefficient for Cu element is determined by the age hardening characteristics of Cu element in aluminum alloys, and is calibrated through aging treatment tests. It is dimensionless. The mass fraction of Cu element is dimensionless. The reference mass fraction of Cu is dimensionless. The fatigue deviation correction factor is determined by the reversibility of fatigue damage in aluminum alloys, calibrated through fatigue repair tests, and is dimensionless. This is the baseline value for the cumulative number of fatigue events, expressed in times. The cumulative number of fatigue events is expressed in times. This is the service damage correction factor; This is the smelting efficiency control coefficient.

4. The method for precise control of automotive aluminum alloy gradient recycling based on edge computing according to claim 3, characterized in that, The adjustment strategy of the real-time feedback unit includes parameter acquisition frequency adjustment and smelting parameter adjustment. The parameter acquisition frequency is adjusted so that when the gradient matching coefficient... When the value is less than 0.6, the real-time feedback unit increases the sampling frequency of the parameter acquisition sub-node. When the gradient matching coefficient is less than 0.6, the sampling frequency of the parameter acquisition sub-node is increased. When the value is greater than 0.8, the real-time feedback unit reduces the sampling frequency of the parameter acquisition sub-node; the smelting parameters are adjusted to match the smelting efficiency control coefficient. When the value is less than 0.5, the real-time feedback unit increases the furnace temperature or extends the holding time.

5. The method for precise control of automotive aluminum alloy gradient recycling based on edge computing according to claim 4, characterized in that, The edge computing node cluster also includes data storage sub-nodes, which are used to store the basic parameters, multi-dimensional recycling parameters, and their corresponding data storage sub-nodes. , , The data storage sub-nodes adopt a distributed storage architecture, and each parameter acquisition sub-node corresponds to a local storage module. The storage period of the local storage module is consistent with the sampling period of the parameter acquisition sub-node.

6. The method for precise control of automotive aluminum alloy gradient recycling based on edge computing according to claim 5, characterized in that, It also includes the deployment of an early warning unit, which monitors in real time. , and The numerical change, when Less than 0.6 Less than 0.5 or When the value is less than 0.4, the early warning unit issues a corresponding service damage warning, smelting efficiency warning, or gradient matching warning. The early warning unit transmits the warning information synchronously with the remote monitoring platform through the local audio-visual module of the edge computing node group.

Images